Sounded Walks the Ethical Tightrope of AI Narration
Elaine Bateman
The intersection of artificial intelligence and the realm of audiobook narration has ushered in a wave of innovation, but it also brings with it a host of ethical considerations that demand careful examination. As the capabilities of AI voice cloning evolve, the ethical landscape of replicating the voices of audiobook narrators has become a focal point of discussion.
One primary concern revolves around the potential impact on the livelihoods of human narrators. The advent of AI voice cloning poses the question of whether automating narration with synthetic voices could diminish opportunities for human narrators in the audiobook industry. Striking a balance between technological advancement and safeguarding employment opportunities within creative fields is a complex ethical minefield.
However, a new audiobook platform, Sounded, hopes its unique approach to the projects will allow it to dance through that minefield unscathed.
We met with Sounded’s Chief Legal Officer Jason Kelly to learn about the new platform and to talk ethics around AI audiobook narration.
What is Sounded?
Sounded (https://authors.sounded.com) is a real-time audiobook production platform, principally for self-published authors and smaller publishing houses, that allows authors or publishers to create an audiobook in real time. Users have a choice of over 824 narrators that are available across sixty languages and 162 dialects. But Sounded is also attempting to balance the scales for narrators by offering a service with real voice artists through a program called True Voice.
“We work directly with narrators to create a digital replica, which is effectively a digital version of their vocal characteristics,” Kelly says. “We license the vocal characteristics from the narrator or voice artist with consent. Any audiobook projects that are narrated using the digital replica are also consented to by the voice artist or narrator. For example, they may not want their voice to be associated with Spicy Romance, or there could be a political connotation that they're not happy about. They have control and can say ‘yes’ or ‘no’ [to a project]. Outside of that, what's very, very important is that the voice artist also receives a commission, even though it's a digital replica that's being used for narration. And again, we're here to work with technology in a responsible manner and work with voice artists to produce the best audiobooks that we can.”
Building a regulated business model is important to Sounded, Kelly explains. “We work on a model of what's known as three C’s, which is consent, control, and commission,” he says. “As we've built a technology-based platform, we've always respected intellectual property rights. We've ensured that we don't use, and haven’t taken, any copyrighted text into our platform. We've also never taken in any audio files or samples from the public domain. As we produce audiobooks with the digital replicas on the platform, the voice artist also has control over the process. They get to decide whether they want their vocal characteristics and digital replica associated with a particular text.”
Kelly believes that the business model of the three C’s works across all AI tech business companies. “Anyone creating an AI product with consent, control, and commission is building a business with a good foundation,” he says.
What does it cost?
Sounded has three levels of membership: the basic tier at $99, the exclusive tier at $249, and the elite tier at $999. At the basic tier, users can select from all 824 narrators. This membership tier allows you to produce an audiobook and distribute it on the platform exclusively for three years, where you set the price point. Additionally, you can offer it for sale in any region you choose. The author receives a royalty payment of 40 percent per sale.
The exclusive tier requires one-year exclusivity with Sounded for the distribution of the audiobook itself but has a 60 percent return to the author or publisher. However, the highest tier, labeled “elite,” allows authors to take audiobook files straight from the platform and distribute them on other platforms. If distributing the audiobook through Sounded, elite users receive a royalty return of 80 percent on a non-exclusive basis. The elite tier also has access to the True Voice platform, meaning users can engage the services of a real voice artist whose digital replica is available on the platform.
How can AI audiobooks be made with ethics in mind?
As with other areas of the publishing world, the advent of AI-narrated audiobooks in recent years have raised plenty of concern. Narrators have pushed back against the use of their voices without their consent to train and produce artificially generated works, especially in cases where the projects may include adult content. On the other hand, authors in support of AI have emphasized the lower cost and ease of production of AI audiobooks, and all big five distributors—Apple Books, Barnes & Noble, Google Play, Kobo, and Amazon—now allow for their creation and/or distribution.
Kelly says one of the biggest areas of debate relates to narrators’ consent to their voices being used for a project. “You have to think of levels of consent. If somebody loses mental capacity at some point, they're no longer able to consent. Equally, if they're deceased, they're no longer able to consent.” As a result, Sounded ensures within its contract all digital replicas on the platform are from narrators who can consent to the use of their voice within their contract, and generated voices on Sounded’s platform are trained only on copyright-expired work.
“There's businesses out there that have different models around ethics and how that works,” Kelly says. “But as a business, we've decided to adopt that particular approach based on advice we've received and also our current beliefs on how to adopt a sensible, regulated position and business model.”
What’s next for the platform?
Sounded launched only a year ago, in April 2023, but Kelly says the program already has new features in the works. “We have a feature called Multicharacter coming later in 2024. It will allow any author to select multiple voices within a book. If you have a novel with many speaking parts, you can have different voices assigned to those speaking parts in the book. The enhanced true voice products are not available right now as a self-service basis, but we have a number of voice artists that we have those digital replicas licensed for. They will all be added to our platform during 2024.”
If you’re interested in finding out more about Sounded, the company will present at the London Stories Festival in September, Frankfurt Book Fair in October, and at book fairs in Bologna and Madrid, or you can contact them at [email protected].
Elaine Bateman